An Adaptive Crocodile Optimization Algorithm Based Deep Elman Recurrent Neural Network for Channel Estimation With Hybrid Precoder in MIMO-OFDM System

Due to the massive usage of smartphones, frequent usage of the IoT, and wireless visual streaming services, data traffic in the wireless network and data explosion has increased over the next years. System modeling and channel estimation are the two main challenges while designing the wireless 5G MIMO communication system. A 2 × 2 MIMO-SFBC system is proposed to enhance the spectral efficiency and capacity of wireless communication systems by exploiting spatial diversity and frequency diversity. The SFBC coding technique gives a low bit error rate (BER) and high signal-to-noise ratio (SNR). Channel modeling and channel estimation are very difficult tasks in the complex propagation characteristics of highly dynamic channels. This paper proposes an improved ERNN-LSTM network to enhance the accuracy and efficiency of channel modeling and estimation in wireless communication systems. Initially, a least squares estimator is employed to obtain an initial estimate of the historical channel responses of a pilot block. These initial estimates are subsequently utilized to train an Elman recurrent neural network (ERNN). The weights of the ERNN's channel parameters are optimized using the Adaptive Crocodile Algorithm. Simulation results show that the proposed ACO-DERNN method achieves a BER of 10⁻⁵ at 30 dB SNR, outperforming conventional methods.